Effects of a 14-day social media abstinence on mental health and well-being: results from an experimental study

The study examines whether abstaining from social media for 14 days affects mental health and well-being outcomes, including depression and anxiety symptoms, loneliness, fear of missing out (FoMO), problematic smartphone use (PSU), screentime, and body image. Given widespread social media use and concurrent increases in mental health problems among young adults, establishing causal effects is challenging with predominantly cross-sectional evidence. The authors leverage longitudinal and experimental methods to infer potential causal effects of social media abstinence. Guided by frameworks such as Compensatory Internet Use Theory (CIUT) and the I-PACE model, the study tests: (H1) PSU positively associates with screentime, depression/anxiety, and FoMO; (H2.1) screentime positively (weakly) associates with depression/anxiety, FoMO, and loneliness; (H2.2) screentime negatively associates with body image; (H3) screentime decreases under abstinence; (H4) depression/anxiety and PSU decrease under abstinence; (H5) body image improves under abstinence. Exploratory questions address daily change patterns in FoMO (RQ1) and loneliness (RQ2), and the stability of effects 14 days post-intervention (RQ3).

Prior research links PSU to adverse outcomes including poor sleep, academic/work impairment, musculoskeletal issues, and psychopathology (depression, anxiety, FoMO). Objective smartphone metrics show mixed associations with depression, and self-reported and objective use correlate only moderately. Problematic social media use (PSMU) shares features with PSU and has been associated with depression, anxiety, stress, cognitive failures, and sleep problems; longitudinal work suggests bidirectional links with depression and associations with insomnia and other outcomes. Reviews of social media abstinence interventions often report small-to-moderate effects: decreased screentime, reduced anxiety/stress, improved sleep and well-being, with mixed findings for FoMO and loneliness; effects may occur even with reduced (not full) abstinence. Body image concerns are linked to social media via exposure to idealized images and social comparison; different screen activities may differentially relate to body dissatisfaction. Theoretical accounts (CIUT, I-PACE) posit that traits and affective/cognitive processes shape risk for excessive use and related psychopathology, implying that abstinence could alter these processes and improve outcomes.

Design and procedure: A longitudinal experimental study ran October 2022–February 2023 using SurveyCoder. After baseline assessment and consent, participants were randomized into four arms: (1) control (use as usual), (2) social media abstinence (deinstall Instagram, Twitter, Facebook, Snapchat, TikTok), (3) gaming abstinence (deinstall all gaming apps), (4) combined abstinence (social media + gaming). Participants received daily survey links at 4 pm during a 14-day intervention, followed by an end-of-intervention survey and a follow-up 14 days later. Participants could otherwise use smartphones normally; app deinstallation was not monitored; other devices were not restricted. Compensation was 10€ or course credit. Sample: Initial N=196 with combined datasets. After exclusions (non-users in experimental arms), n=165 remained. Due to 83.6% female composition and a focus on body image, analyses in the main manuscript were restricted to females. The gaming (n=21) and combined-abstinence (n=31) groups were excluded from main analyses, leaving n=86 females: control n=35, social media abstinence n=51. Groups did not differ in age, education, or occupational status. Measures and timing: - Baseline, end-of-intervention, follow-up: TS FoMO (trait and online-specific state); PHQ-4 (depression/anxiety); SAS-SV (PSU); UCLA 3-item Loneliness; BIAS-BD (body image dissatisfaction via actual–ideal body discrepancy); MBSRQ-AS subscales (appearance evaluation, appearance orientation, body area satisfaction, overweight preoccupation, self-classified weight); smartphone screentime (7-day average from device screentime feature). Baseline covariate: Fear of COVID-19 (FCV-19S). - Daily (14 days): Single-item FoMO (FoMOsf), UCLA 3-item Loneliness, PHQ-4 daily, and daily screentime. Links sent daily at 4 pm. Data handling: Screentime was self-entered from device screentime summaries (converted to minutes). No OS-specific distinctions. Plausibility checks applied. Statistical analysis: Conducted in R 4.1.3 with lme4 and lmerTest. Baseline descriptive statistics and Holm-corrected correlations. Multilevel models for outcomes measured at three time points used spline models with a knot at end-of-intervention to estimate pre- and post-knot trends (Model 1: time trends only; Model 2: adds covariates including FCV-19, baseline PSU, and baseline BID for MBSRQ-AS outcomes, plus group effects with control=0, abstinence=1). Random intercepts (and random slopes where possible) modeled interindividual variability. Daily outcomes (FoMO, loneliness, screentime, PHQ-4) were modeled with linear, quadratic, and cubic time trends; models compared via AIC, BIC, and likelihood ratio tests. Covariates for daily models included baseline outcome value, baseline FCV-19, PSU, and BID. Ethics approval: University of Ulm EC (252/22). OSF repository: https://osf.io/qdp8r/.

Baseline associations (H1–H2): - PSU correlated positively with depression/anxiety and FoMO (state>trait), aligning with H1; PSU was not significantly related to screentime. - Screentime showed weak positive associations with depression/anxiety, FoMO, and loneliness, and weak negative associations with appearance evaluation/body area satisfaction and positive with appearance orientation; effects were small and mostly nonsignificant, broadly consistent with H2.1–H2.2 expectations of weak links. Three-timepoint multilevel spline models: - Screentime: No overall change in Model 1. In Model 2, abstinence vs. control showed a greater decrease from baseline to end of intervention (T1*Group b = −38.905 min, t = −1.685; one-sided p = .047), supporting H3. Changes from end to follow-up were nonsignificant, suggesting stability. - PSU: Significant pre-knot decrease overall (T1 b = −4.023, t = −5.583, p < .001); no group differences, so H4 for PSU not supported as an abstinence-specific effect. - Depression/anxiety (PHQ-4): Overall pre-knot decrease (T1 b = −0.60, p = .013) in Model 1; no group differences in Model 2; H4 not supported for PHQ-4. - Loneliness: Decreased from baseline to end (T1 b = −0.66, p < .001) without group differences. - FoMO: State and trait both decreased from baseline to end (state T1 b ≈ −0.22 to −0.52; both p < .001 in Model 1); no differential group trends. Trait FoMO showed lower values at the knot for the abstinence group but similar change rates. - Body image: BID showed a greater reduction in the abstinence group from baseline to end (T1*Group b = −0.95, t = −1.900; one-sided p = .029), supporting H5 (small effect). Appearance evaluation increased overall (pre-knot), body area satisfaction showed a marginal increase in Model 1; overweight preoccupation decreased in the control group in Model 2; no group differences for these. Daily multilevel models (14 days): - Screentime: Linear/quadratic time trends were weak; interactions suggested different patterns for the abstinence group (near-significant interactions) consistent with overall decrease observed in spline models. - Loneliness: Best fit with a cubic trend over days; no group differences (RQ2). - FoMO: Best fit with a cubic trend; significant linear, quadratic, and cubic components; no group differences (RQ1). - Daily PHQ-4: Interaction between day and abstinence group (b ≈ 0.06, p < .05) indicated differing trajectories; control showed a small decrease (p ≈ .067). Overall effects were small. Stability to follow-up (RQ3): Most outcomes showed little to no change from end of intervention to 14-day follow-up, indicating short-term stability of observed changes (e.g., screentime and BID).

The experimental abstinence design, combined with longitudinal modeling, provides evidence that a 14-day social media break can reduce overall smartphone screentime and modestly lower body image dissatisfaction compared to continued use. Broader improvements (decreases) in PSU, depression/anxiety, FoMO, and loneliness occurred across both groups, suggesting common temporal trends or behavioral intentions beyond the assigned intervention. The lack of group differences for most mental health outcomes indicates that simply removing selected social media apps may be insufficient to produce distinct changes in these domains within two weeks, possibly due to compensatory use of other smartphone activities or the role of dispositional factors (e.g., FoMO as a trait) in the I-PACE model. Daily analyses revealed cubic fluctuation patterns in FoMO and loneliness, underscoring dynamic day-to-day variability rather than linear change. Together, these findings nuance causal interpretations: abstinence specifically reduced screentime and BID but did not uniquely alter general psychopathology measures in this timeframe. They point to individual differences (e.g., motives of use per CIUT, trait vulnerabilities) as potential moderators of intervention response.

Using a longitudinal experimental approach, the study shows that a 14-day social media abstinence leads to reduced smartphone screentime and modest reductions in body image dissatisfaction, with effects stable over a 14-day follow-up. Other outcomes (PSU, depression/anxiety, FoMO, loneliness) improved over time irrespective of group, suggesting broad trends or intentions to reduce use, and potential compensatory behaviors. The work advances understanding of how short-term abstinence affects body image and daily fluctuations in FoMO and loneliness. Future research should: include objective usage tracking and verification of abstinence; assess problematic social media use (PSMU) and motives of use; examine moderators (e.g., baseline psychopathology, trait FoMO, gender); test varying abstinence dosages (reduction vs. abstinence) and longer follow-ups; and extend to more diverse and larger samples, including balanced gender representation and clinical populations (e.g., eating/body schema disorders).

- Sample restrictions and power: Main analyses included only female participants and excluded the gaming and combined-abstinence groups, yielding small group sizes and reduced statistical power; some effects were directionally consistent but nonsignificant. - Measurement limitations: Problematic social media use (PSMU) was not assessed; screentime was self-entered from device summaries (potential transfer errors) and not objectively logged; no monitoring or verification of app deinstallation or use on other devices. - Intervention specificity: Abstinence targeted selected social media apps only; participants could compensate with other smartphone activities, potentially diluting effects on mental health outcomes. - Unmeasured moderators/mediators: Motives and purposes of smartphone/social media use were not assessed; individual differences (e.g., escapism, trait vulnerabilities) may moderate effects. - External events and generalizability: Conducted during the COVID-19 period (FCV included as covariate) and among predominantly students/young adults, limiting generalizability. Only a single short follow-up (14 days) was included, limiting conclusions about long-term maintenance.